Data representation for sequential models
Timesteps add depth to the data, making it a 3D structure. In the context of sequential data, each “unit” or instance of this dimension is termed a “timestep.” This is crucial to remember: while the dimension is called “timesteps,” each individual data point in this dimension is a “timestep.” Figure 10.4 illustrates the three dimensions in data used for training RNNs, emphasizing the addition of timesteps:
Figure 10.4: The 3D data structures used in RNN training
Given that the concept of timesteps is a new addition to our exploration, a special notation is introduced to represent it effectively. A superscript enclosing a timestep in angle brackets is paired with the variable in question. For example, using this notation, 
and 
represent the value of the variable stock_price at timestep t1 and timestep t2, respectively.
The choice of dividing data into batches, essentially...
